Titulo Estágio
Algorithmis for Patient stratification and alarms generation
Área Tecnológica
Informática Médica
Local do Estágio
DEI
Enquadramento
Most therapies are currently prescribed empirically. The philosophy behind personalized medicine is that every patient has a unique biology and physiology that should be reflected in the choice of therapy, thus resulting in an improved treatment outcome. Currently, the implementation of personalized medicine is considered a stepwise process where stratification of patients into subgroups is the first important step. In this thesis proposal, we aim the development of a patient stratification and alarm model to support diagnosis and clinical decision and to derive practical and effective alarms for clinicians. It mainly involves the development of a knowledge base (KB) on patient stratification and alarms generation using several types of available clinical knowledge and, eventually, knowledge mined from databases.
Medical knowledge is as extensive as it is sparse and discontinued. It can exist implicitly, in the form of raw patient data or physician experience, or explicitly in clearly defined clinical rules and diagnostic protocols. Moreover, this knowledge might not be consistent; it may present different levels of certainty, granularity and completeness. Given the heterogeneity of the medical knowledge sources, two main problems have to be solved: (1) knowledge representation and (2) knowledge integration or fusion. In this proposal we shall tackle both problems in a biomedical context.
Objetivo
The goal is to develop and implement a patient stratification methodology involving the physician in the stratification loop by integrating data-driven and knowledge-driven (physician and available literature) methodologies. This methodology will enable to stratify patients according to their care needs as well as to identify indication of possible causes.
In this project we propose to develop a knowledge base representation and fusion framework able of integrating explicit and implicit knowledge in the context of patient. In a first step, the knowledgebase for deriving alarms and patient stratification will be based on existing clinical knowledge such as clinical protocols and literature survey. In a second step, this knowledgebase will be extended using implicit expert knowledge captured, for instance, through questionnaires and interviews. Finally, in a third step, the KB might be refined and extended using a data driven approach based on data collected throughout the project in small scale studies or from existing databases.
This framework will be applied to define a suitable strategy for patient stratification according to their treatment needs based on their condition and predicted evolution. The strategy will be developed in order to enable the detection of possible cause, hence enhancing diagnosis and decision support to the physician.
Using a similar approach a knowledgebase for alarm detection and prioritization will be developed in the context of the project.
In a first step, the knowledgebase for deriving alarms and patient stratification will be based on existing clinical knowledge such as clinical protocols and literature survey. In a second step, this knowledgebase will be extended using implicit expert knowledge captured, for instance, through questionnaires and interviews. Finally, in a third step, the KB might be refined and extended using a data driven approach based on data collected throughout the project in small scale studies or from existing databases.
Plano de Trabalhos - Semestre 1
1 – Domain familiarization
2 – State of the art on medical knowledge representation strategies
3 – State of the art on patient stratification using existing protocols and expert capturing tools
4 – Development of a knowledge base capturing tool
5 – Writing of the thesis proposal
Plano de Trabalhos - Semestre 2
6 – Research and development of the DSS framework based on expert knowledge (knowledge drive approach)
7 – Research and development of the DSS framework based on data driven approach
8 – Application to a specific clinical problem
9 – Writing of the thesis
Condições
A realizar no DEI
Orientador
Paulo de Carvalho
carvalho@dei.uc.pt 📩